Function Stack Duration difference

Vladimir Slavinsky

Typically, most function stacks have timing as the sum of all operations. But during the day there are such discrepancies. What could be the reason for this?

Michael Udinski

One potential thing could be due to large outputs in the response which requires time to render the data in the response. Overall it doesn't look like performance has taken a significant hit. How much data are you returning?

Vladimir Slavinsky

Thanks for your interest in the issue.

To begin with, I do not agree that "it did not suffer much." 1.3 seconds instead of 0.2 seconds is a x6 difference.

In answer to your question, the amount of data is as follows. These are 2 lists from the same table with 15 columns, mostly integer and long, the rest are strings, usually empty. It often not more than 10KB.

The number of rows in the 1st query is from 0 to 200, but most often 20-50. There are no addons.

The number of rows in the 2nd request is from 0 to 2. There are 2 small addons.

But the main thing is not the amount of data, but the fact that the same query with the same result (20-50 rows), executed at different time intervals, we have different execution times. Often I get around 0.2 seconds per queries and the stack execution time is 0.3 seconds. It's not bad. But sometimes it can be same 0.2 seconds per queries and the stack execution time is over 1 second. At the same time, the load on the API and database does not exceed 2-3% (no users). That's what's weird.

Vladimir Slavinsky

3 requests from same user in the interval of 5 minutes.

This time I specifically counted, in all 3 cases the response had 69 rows

Load API - 1%

Load Database - 1%

We are ready to promote our product but we are afraid to fail

😢

Please help. We want to stay with Xano

Sean Montgomery

@Vladimir Slavinsky it would be useful if you run the same request at least 10x for 5 minutes straight - so 50x total. Xano handles caching internally and a cache item could be getting created which could affect performance for a brief moment. There may be something else, but it is hard to know without more detail on your side.

Seeing those 50 requests would help shed some light as to what is going on and let us see if there is an issue.

I would also like to point out that we recommend at least Scale for any product being released if you are expecting any type of traffic higher than basic testing. This isn't to say we are ignoring your screenshots above - we definitely want to get to the bottom of that as well.

Vladimir Slavinsky

@Sean Montgomery

Thanks for your interest in the issue.

Following your advice, I did the test. This time I got 61 entries each time.

API load was 3%

Database load was 1%

Everything worked well.

A little later, I ran a more aggressive test at 2-3 second intervals. In addition, I swiched the same query with queries for adding and deleting data in the same table, which in my opinion should complicate the task of indexing and I assumed that I would get increased delays.

API load was 15%

Database load was 1%

But I got exactly the same result.

As you can see, the tests were carried out at night, and I received large delays (1st post) during the day. I think I need to repeat the same test in the daytime.